The Human Needs Index (HNI) was developed through a rigorous methodology with the goal of informing nonprofit leaders and policymakers about the degree of poverty-related need and measuring the effectiveness of response to need, over time, in specific communities and across the country. To achieve this goal, a standardized index was constructed.
Constructing the HNI was an iterative process that employed empirical evidence, statistical methodologies, and expert consultation. We began this process by developing a conceptual framework for analyzing poverty by identifying appropriate line-item service variables that represented essential basic needs. The selection of variables was based on careful consideration of the academic literature and theoretical constructs associated with measuring poverty and human needs. Overall, we tested more than 450 combinations of 230 organizational service variables to create the HNI – at the national, regional, and state level. We initially selected 21 material assistance and personalized service variables representative of basic human need – that is, the delivery of food, clothing, shelter, or health/well-being services.
Initial tests of these variables were conducted at the national level and by year, although subsequently, data were disaggregated to test variables at the service center site, county, state and territorial levels, and by month. In testing the variables’ ability to measure human need, we relied on external governmental measures of poverty, including the poverty rate, unemployment rate, and the Supplemental Nutrition Assistance Program (SNAP) benefit usage. The statistically significant associations demonstrated with these government measures provided guidance in selecting which variables to include in preliminary modeling.
Next, the 30-member team of statisticians, program officers, economists, and National Advisory Board members from The Salvation Army and Indiana University Lilly Family School of Philanthropy engaged in detailed discussions about which variables dedicated to food, clothing, shelter, or health/well-being services were collected in all states and across all four regions during each year. As a result, three variables (Meals Provided, Clothing Provided, and Lodgings Provided) were initially selected for preliminary testing against the external government measures, individually and together as a test model. The results revealed positive and significant correlations. Thus, additional variables representing these basic needs were added for further model testing.
In the final iterations of creating the HNI, using the governmental measures as benchmark validation and the 21 material assistance and personalized service variables representative of basic human need, we employed three approaches to weighting variables to determine appropriateness for inclusion or exclusion in the final national and state-level models. The most parsimonious model included three variables representing food, shelter, and health/well-being services. The weighting for this model was based on the geometric mean of the included variables. The second technique was the most analytical and pragmatic, in that it used both theoretical and empirical justifications for the weighting and selection of variables. The initial testing of this approach included five variables representing food, shelter, clothing, and health/well-being services. The third and final approach was also an analytic model that included all 21 variables. This process concluded with the presentation of the six strongest models derived from these distinctive approaches.
Ultimately, the second approach using Principal Components Analysis (PCA) was used to build the HNI because it allowed for both intuitive variable selection as well as statistical confirmation of individual variables’ utility in the overall model measuring human need. Subsequently, to ensure the strongest, most appropriate model was selected, further discussions about, and statistical testing of, the second approach were conducted. This concluding phase testing necessitated the retention of four variables, the removal of one variable, and the addition of three variables. The final model, therefore, includes seven (7) line-item variables and demonstrates strong correlations with benchmark data. Another important strength of the final model used to construct the HNI is that it is built from variables representing essential aspects of human need that are measured by The Salvation Army consistently across time and region.
